Deprecation warning of prob_correct()

[Jiashun97]

Hi Max,

I hope this message finds you well. I received this warning many times when I used the LossByMeans class for the customized loss function. https://pyddm.readthedocs.io/en/stable/cookbook/loss.html#loss-means

Warning: This function (prob_correct()) is deprecated and will be removed in a future version of PyDDM. Please use Solution.prob('correct') instead.

However, I didn't use the prob_correct() function and I think the problem is related to the mean_decision_time() function. I used the version 0.8.1. Could you figure out why this is the case? Thanks for your help.

Best regards,
Jiashun

[mwshinn]

Hi Jiasun,

This will not be a problem for fitting, but if you'd like the error message fixed, please post all the information described in: https://pyddm.readthedocs.io/en/latest/contact.html#how-should-i-ask-for-help

That being said, LossByMeans can give pretty bad fits and I wouldn't recommend using it, it is more for illustration of how to build a loss function.

[Jiashun97]

Thanks for your reply. I am now trying to implement Quantile Maximum Likelihood (QML) as the loss function. When I fitted an Ornstein-Uhlenbeck model to datasets with fast RTs using likelihood as the loss function, I noticed that it tends to over-emphasize the RT distribution shape, leading to poor choice proportion predictions. Specifically, I found that the drift rate is small while the self-excitation (negative leak) is high, making the decision process overly deterministic. Do you think there is another way to make the model prediction better instead of changing the loss function?